What Interdisciplinarity Really Means (to me)

I was trained in physics. I was taught in college that therefore, physics is the only thing I should really go to much trouble to learn. After all, being good at physics allows you to do good science. Stands to reason?

Well... I disagree.

At least in astronomy (observational, especially), learning about other fields is just as important. Astronomers are set in their ways--we still fit lines with heteroscedastic errors in both variables by linear weighted-least-squares (headdesk)--and astronomers are famously bad at statistics. We need to encourage (and yes, even force) students to take methods-oriented classes (i.e., 500-level and above) in disciplines like computer science, statistics, and mathematics. That's the only way to guarantee exposure to modern techniques used in actual research.

An example: suppose you want to filter out cosmic rays from an image. The old-old way is to smudge them out by hand. No good for gigapixel-size images. Nowadays people use Laplacian edge detection, which is horribly temperamental. If you're keeping up-to-date on the facial recognition scene, though, you may have heard that people are working on smile detection, and some are using local binary patterns (basically, simultaneous measurements of an pixel's value, compared to its neighbors--all smushed into a feature with 59 possible values). The upshot is that you get recognizable (and illumination-invariant) values near where there are edges, and filter out absolutely everything else. No pesky over-filtering. No fuss, no muss.

Modern methods work. They're backed up by recent literature, they're in textbooks, and they're used. Most of all, they're beginning to be included in software libraries.

Interdisciplinarity, to me, is the awareness of what is going on in other fields, and the willingness to take a little extra time to apply that knowledge.